>> I noticed this same thing yesterday when benchmarking XFree86 3.2. It
>> does look like it's potentially a kernel problem; there seems to be a
>> much higher proportion of system (kernel) time to user time while X
>> benchmarks like xbench and x11perf are running than seems reasonable,
>> about 50/50 if xosview is believable. The benchmarks themselves seem to
>> draw in a way similar to what would happen if they were being run with a
>> high CPU load on the system, with jerkiness in the drawing.
I noticed the same slowdown with 2.0.24 & X.
I'am running those tests on a P133/64M/Matrox Millenium/Accel-X 1.2.
All benchmarks are under 8bpp. The machine is not connected to a
network, except for dial-up IP using ppp, so I have no networking card
drivers loaded.
| | | | | | | | |
machine |p| comm | line | fill |blt | text | arc |cmplx |xstones
------------+-+------+-------+-------+------+-------+--------+------+-------
Linux 2.1.7 |8| |1160658|1031891|363361|2101001|37697225|859477|918021
------------+-+------+-------+-------+------+-------+--------+------+-------
Linux 2.0.23|8| |1091383|1018356|383735|2056314|37163409|670065|904880
------------+-+------+-------+-------+------+-------+--------+------+-------
Linux 2.0.24|8| | 615362| 522026|237848|1113751|19480827|820392|550898
Note: I have trimmed the results a bit, to make them fit on a line :)
Note2: 2.0.23 and 2.0.24 are compiled with the fpu-memcpy patch.
For additional info, I will add some lmbench results. I hope, this may
help to shade some light on.
L M B E N C H 1 . 0 S U M M A R Y
------------------------------------
Processor, Processes - times in microseconds
--------------------------------------------
Host OS Mhz Null Null Simple /bin/sh Mmap 2-proc 8-proc
Syscall Process Process Process lat ctxsw ctxsw
--------- ------------- ---- ------- ------- ------- ------- ---- ------ ------
darkstar Linux 2.1.6 133 1 0.9K 7.2K 20K 148 7 12
darkstar Linux 2.0.23 133 3 0.9K 7.0K 19K 167 7 10
darkstar Linux 2.1.7 133 1 1.0K 7.2K 20K 149 6 10
darkstar Linux 2.0.24 133 2 0.8K 6.9K 19K 168 6 10
*Local* Communication latencies in microseconds
-----------------------------------------------
Host OS Pipe UDP RPC/ TCP RPC/
UDP TCP
--------- ------------- ------- ------- ------- ------- -------
darkstar Linux 2.1.6 24 175 359 233 449
darkstar Linux 2.0.23 31 182 347 241 466
darkstar Linux 2.1.7 24 160 328 217 429
darkstar Linux 2.0.24 30 191 382 231 491
*Local* Communication bandwidths in megabytes/second
----------------------------------------------------
Host OS Pipe TCP File Mmap Bcopy Bcopy Mem Mem
reread reread (libc) (hand) read write
--------- ------------- ---- ---- ------ ------ ------ ------ ---- -----
darkstar Linux 2.1.6 43 16.7 41.5 66.5 43 40 73 83
darkstar Linux 2.0.23 35 15.1 46.0 64.8 43 40 73 83
darkstar Linux 2.1.7 44 16.2 41.5 66.1 43 40 73 83
darkstar Linux 2.0.24 35 16.1 46.4 65.1 43 40 73 83
Memory latencies in nanoseconds
(WARNING - may not be correct, check graphs)
--------------------------------------------
Host OS Mhz L1 $ L2 $ Main mem TLB Guesses
--------- ------------- --- ---- ---- -------- --- -------
darkstar Linux 2.1.6 132 7 108 182 426
darkstar Linux 2.0.23 132 7 101 182 428
darkstar Linux 2.1.7 132 7 101 182 430
darkstar Linux 2.0.24 132 7 108 182 426
-- # /AS/ mailto:alex@darkstar.ping.at God save the screen ! # # http://members.ping.at/alexa/ powered by Linux 2.0 # ------------------------------------------------------------------------